Archives - Page 4

• 

  Real-Time Anomaly Detection in Streaming Sensor Data Using LSTM Autoencoders

  Vol. 2 No. 10 (2025)

  In the era of Industry 4.0 and the Internet of Things (IoT), billions of connected sensors continuously generate large volumes of real-time data streams. This sensor data is vital for decision-making in domains such as industrial automation, predictive maintenance, and critical infrastructure monitoring. However,  these  systems  are  susceptible  to  irregularities  caused  by  sensor  faults,  environmental disturbances, or cyber intrusions. Detecting such anomalies in streaming data is challenging due to the velocity, volume, and evolving nature of the streams.
  This research introduces a Real-Time Anomaly Detection Framework using Long Short-Term Memory  (LSTM)  Autoencoders,  designed  specifically  for  processing  continuous,  high-velocity sensor data. Unlike conventional models that rely on static, offline data, the proposed model learns temporal dependencies dynamically and adapts to new patterns using an incremental sliding window mechanism. The LSTM Autoencoder reconstructs normal time-series sequences, and any significant deviation between input and reconstruction indicates an anomaly. The framework integrates with streaming platforms like Apache Kafka and Apache Flink, enabling low-latency inference.
  Experimental evaluations on real-world industrial datasets demonstrate that the proposed approach achieves  superior  precision  (0.96) and  F1-score  (0.94) while  maintaining  latency  below  100 milliseconds. The system adapts to changing patterns in real time, offering robustness against concept drift. This work contributes toward developing intelligent, adaptive, and explainable anomaly detection systems applicable to diverse real-time environments such as smart manufacturing, IoT-enabled grids, and autonomous systems.

• 

  Deep Learning-Powered Automated Detection of Abnormalities in Chest X-Rays

  Vol. 2 No. 04 (2025)

  Abstract

  Medical imaging plays a crucial role in diagnosing various diseases and abnormalities within the human body, with chest X-rays being one of the most commonly used modalities. In recent years, deep learning techniques have shown remarkable promise in automating the analysis of medical images, including the detection of abnormalities in chest X-rays. This project aims to explore the application of deep learning algorithms, particularly convolutional neural networks (CNNs), for the automated detection of abnormal findings in chest X-rays. The project will involve the collection and preprocessing of a diverse dataset of chest X-ray images, encompassing both normal and abnormal cases. Subsequently, deep learning models will be trained, validated, and fine-tuned using the collected dataset to accurately classify chest X-rays as either normal or abnormal based on the presence of various pathologies such as pneumonia, lung nodules, or pleural effusion. The performance of the developed models will be evaluated using standard metrics such as accuracy, sensitivity, specificity, and area under the receiver operating characteristic curve (AUC-ROC). The outcomes of this project aim to contribute to the advancement of computer-aided diagnosis systems in healthcare, potentially aiding clinicians in making more accurate and timely diagnoses, thus improving patient outcomes.

  Index Terms

  Medical Imaging, Chest X-rays, Deep Learning, Convolutional Neural Networks (CNNs), Automated Detection, Abnormal Findings, Dataset Collection, Preprocessing, Pathologies, Pneumonia, Lung Nodules, Pleural Effusion, Performance Evaluation, Accuracy, Sensitivity, Specificity, Area Under the Receiver Operating Characteristic Curve (AUC-ROC), Computer-Aided Diagnosis Systems, Healthcare, Patient Outcomes.